Customer data glossary
Identity graph
By Art Kukharenko · Published · Updated
Browse learn glossary resourcesAn identity graph is a data structure that stores the links between identifiers (emails, device IDs, account numbers) and the person or household they have been resolved to.
Part of the Learn hub and the customer data glossary.
What it means in practice
In graph terms, each identifier is a node and each piece of evidence connecting two identifiers is an edge: a login that used both an email address and a device, a form that supplied a phone number alongside an account number. A connected cluster of nodes represents one person, and the cluster receives a stable person ID that downstream systems use.
Useful graphs keep metadata on every edge: the source system, the time observed, the matching method and a confidence value. That metadata is what makes a merge explainable and reversible, and it lets you answer a regulator or a customer who asks where a link came from.
Graphs differ in provenance. A first-party graph is built from your own logins, transactions and forms. A commercial or cross-device graph is licensed from a data vendor and rests on third-party data, which brings its own questions about consent and accuracy.
A graph database, relational tables or warehouse files can all hold a graph. What matters is that deletion works on the graph itself: a deletion or suppression request has to reach every edge and identifier, not only the profile table, or the person can be re-identified.
Example (insurance): an insurer links a policy number, a portal login and a claims-line phone number. When the policyholder withdraws marketing consent, the graph lets the platform find every identifier tied to that person and suppress them all, rather than only the email address typed into the last form.
Related terms
- Identity resolution: the process that builds the graph.
- Deterministic matching: creates high-confidence edges.
- Probabilistic matching: creates scored edges.
- Third-party data: the source of vendor-licensed graphs.
- Golden record: the profile assembled from a cluster.
Where it matters
- Identity resolution explained: how graphs are built and tested.
- Customer data platform architecture: where the graph sits between ingestion and activation.
Frequently asked questions
Is an identity graph the same as a customer data platform?
No. An identity graph is one component: the store of identifier links. A customer data platform also collects events, builds profiles, enforces consent and activates segments.
Who owns an identity graph?
A graph built from your first-party data and held in your environment is your asset, subject to privacy law. A vendor-licensed graph is licensed, usually with terms on use, retention and what happens when the contract ends. Check the exit terms before depending on it.
Can an identity graph run on-premise?
Yes. It is data plus matching logic, so it can run in your own data centre or cloud account. The requirement is that storage, deletion and audit all sit inside the environment you control.
Back to the customer data glossary or the Learn hub.
Further reading
- Customer Data Glossary: 50 Terms
- Learn about CDP
- CDP Architecture: The 3 Layers
- Deterministic matching — definition
- Golden record — definition
Published: · Updated:
Publication date recorded for the approved release. Editorial updates do not establish legal approval or a customer result.